【作者】 彭玉楼；

【导师】 何怡刚；

【作者基本信息】 湖南大学 ， 电气工程， 2013， 博士

【摘要】 随着云技术的发展，数字化产品在人类生活中的广泛应用，有效保护数字多媒体数据版权成为当今时代的需要。数字水印技术作为有效的手段之一，自上世纪90年代出现以来，得到广大研究人员的高度重视。数字水印技术是将特定信息嵌入到宿主数据中，从而保护宿主数据的安全，其内容涉及到版权保护、内容认证、指纹认证、广播监控、注解标记等方面，但是数字水印技术仍未成熟，在实际应用中还存在诸多问题，如实现数字水印走向商业化的鲁棒性水印研究问题。因此，本文以静态图像作为研究对象，主要在鲁棒性水印的设计方面进行了研究，通过结合图像的特征、视觉模型实现了数字水印在空域和变换域的嵌入和检测，为数字水印技术的实际应用提供了有效的解决方案，并对所提出算法进行详细的性能分析和计算机仿真验证。论文的主要创新点归纳如下：（1）提出了一种基于小波变换和混沌的函数水印算法。针对图像水印的嵌入量对图像透明性的影响，数字水印由用户自定义的函数产生，扩展了水印的生成空间。对灰度图像进行小波变换，在小波变换的中低频域选择水印嵌入点嵌入水印，通过Logistic混沌对嵌入点坐标加密，用户在得到密钥和加密序列后可提取水印，实验结果表明算法对于原始图像的改变小，水印具有不可见性。算法对常见的水印攻击具有较好的鲁棒性，但在噪声攻击中的高斯噪声和泊松噪声提取水印效果不是很理想。为了提高混沌系统的安全性，设计了五维超混沌系统，并数值证明了系统的可行性。（2）提出了基于Contourlet变换和ECC（Elliptical Curve Cryptography）的数字水印算法。在引入中央可信IPP （Intellectual Property Protection, IPP）机构的基础上，提出了基于ECC的公钥数字水印的算法，利用ECC体制的加密解密的优点，在Contourlet小波域中选择避开图像边缘的高频部分，实现水印的嵌入，算法对常见的水印攻击具有较好的鲁棒性，为水印加密体制的标准化提供了有意义的探索。（3）提出了基于压缩感知和混沌的函数数字水印算法。在数字水印的框架中，提高水印的嵌入量是目前解决的主要问题之一，分析了函数水印的稀疏性，利用压缩感知理论可嵌入更多的数据，提高水印容量。为了提高压缩感知重构信号的精度，提出了基于SVD（Single Value Decomposition）的观测矩阵设计算法，证明了算法的有效性，实验仿真结果验证可以提高一维信号的重构精度3-5%，提高二维信号的PSNR值1-2dB；在图像归一化的基础上，提出了基于压缩感知和混沌的函数数字水印算法，实验仿真结果表明算法对于几何攻击和常规攻击具有良好的鲁棒性，可以提高水印嵌入量大约7倍。（4）提出了基于数字水印的学生程序竞赛集训管理系统和基于数字水印的水稻信息管理系统，为数字水印的实际应用提供了有意义的探索。两个系统是数字水印在彩色图像中的应用，虽然彩色图像的数字水印算法总体上是类似于灰度图像的数字水印算法，但是与灰度图像水印算法又有不同之处，主要在于彩色图像的水印嵌入要考虑到人类对彩色的视觉模型，考虑到不同的实际应用，需要进行颜色空间变换。在提取水印时，利用尺度不变特征变换SIFT（Scale InvariantFeature Transform)对图像校正，由于校正需要原始图像，算法是非盲水印算法，两个系统的运行结果表明了系统对于噪声攻击和几何变换具有良好的鲁棒性。更多还原

【Abstract】 With the development of cloud technology, the broad application of digitalproducts in human life, effectively protecting the copyright of the digital multimediadata becomes the need of the present era. The digital watermarking technology is as akind of the effective protection means since the1990s, and gets attached greatimportance to the majority of researchers. The digital watermarking technology is thatspecific information is embedded into the host data, so as to protect the security of thehost data, its contents is related to copyright protection, content authentication,fingerprint authentication, broadcast monitoring, annotations mark; But it is not yetmature, there are still many problems in practical applications, such as robustnessdigital watermarking research towards commercialization. Therefore, in thisdissertation, a static image is taken as the research object, mainly in the robustness ofthe watermark design research. By the combination of the characteristics of the imageand the visual model to achieve the digital watermark in the spatial domain andtransform domain embedding and detection, effective solutions to the practicalapplication are provided, and a detailed performance analysis and computersimulation of the proposed algorithm is given. The main innovations of thisdissertation are summarized as follows:Firstly, a functional digital watermarking algorithm based on wavelet transformand chaos is proposed. Aiming at the influence of image transparency for watermarkcapacity, the digital watermark is generated by a user-defined function to extend thewatermark generation space. Using wavelet transform deal with grayscale images,choosing embedding points of middle-low frequency in wavelet transform, watermarkembedding points is encrypted by Logistic chaos, user can extract the watermark byobtaining the key and encryption sequence. The experimental results show that thealgorithm changes little to the original image, the watermark is not visible, and isrobust to common watermark attacks. In order to improve the security of chaoticsystems, a five-dimensional hyper-chaos system is designed and the feasibility of thesystem is proved adopting numerical method.Secondly, digital watermarking algorithm based on contourlet transform andECC(Elliptical Curve Cryptography) is proposed. ECC public key digitalwatermarking algorithm is given out based on a central trusted IPP (Intellectual Property Protection) institution, the watermark is embedded in contourlet domainavoiding the high frequency of the image edge, and the algorithm has betterrobustness of common watermark attacks, and provides a meaningful exploration forwatermark encryption system standardized.Thirdly, a digital watermarking algorithm based on compressed sensing andchaos is proposed. Currently one of the main issues is improving the embeddedwatermarking capacity in architecture of digital watermarking. More data isembedded to improve the watermark capacity by use of compressed sensing theory,and the sparsity of the functional watermark is analyzed. In order to improve thesignal reconstruction precision a measurement matrix design algorithm is presentedbased on SVD（Single Value Decomposition）, the effectiveness of the algorithm isproved, the simulation results verify that can improve the reconstruction accuracy3-5%for one dimensional signal and PSNR1-2dB for of two dimensional signal;digital watermarking algorithm is putted forward based on compressed sensing andchaos in normalized image, simulation results show that the algorithm has goodrobustness to geometric attacks and conventional attacks, the embedded informationcapacity of watermarking is enhanced about7times.Finally, the student contest management system and rice informationmanagement system based on digital watermarking are proposed, and a meaningfulexploration of the practical application of digital watermarking are provided. Theapplication of digital watermarking is in color image in the two systems, color imagedigital watermarking algorithm is similar to grayscale images overall, but hasdifferent with a grayscale image watermark algorithm, the difference mainly lies inthe color image watermark embedding taking into account the human color visualmodel, and should consider color space transform in the practical application. Whileextracting watermarking, SIFT (Scale Invariant Feature Transform) is used to imagecorrection, the algorithm is non-blind watermarking, and the operating results of thetwo systems shows that the two systems are robust to noise attack and the geometrictransformation.更多还原